Clinical & Experimental Metastasis 21: 159–166, 2004.© 2004 Kluwer Academic Publishers. Printed in the Netherlands.

159

A soybean Kunitz trypsin inhibitor suppresses ovarian cancer cell invasion byblocking urokinase upregulation

Hiroshi Kobayashi, Mika Suzuki, Naohiro Kanayama & Toshihiko TeraoDepartment of Obstetrics and Gynecology, Hamamatsu University School of Medicine, Hamamatsu, Shizuoka, Japan

Received 28 October 2003; accepted in revised form 27 January 2004

Key words: Kunitz, protease inhibitor, soybean, urokinase-type plasminogen inhibitor

Abstract

We have previously reported in a series of papers that a Kunitz-type protease inhibitor, bikunin, suppresses up-regulationof urokinase-type plasminogen activator (uPA) and its specific receptor (uPAR) expression, phosphorylation of ERK1/2and cancer cell invasion in vitro and peritoneal disseminated metastasis in vivo. In the present study, we investigated theeffects of soy bean trypsin inhibitor (SBTI) on the net enzymatic activity of secreted, extracellular uPA, signal transductioninvolved in the expression of uPA and invasion in HRA human ovarian cancer cells. SBTI contains a Kunitz trypsin inhibitor(KTI) and a Bowman–Birk inhibitor (BBI). Here, we show 1) that KTI and BBI were purified separately from soybeans; 2)that neither KTI nor BBI effectively inhibits enzymatic activity of uPA; 3) that uPA upregulation observed in HRA cells wasinhibited by preincubation of the cells with KTI with an IC50 of ∼ 2 µM, whereas BBI failed to repress uPA upregulation,as measured by enzyme-linked immunosorbent assay; 4) that cell invasiveness was inhibited by treatment of the cells withKTI with an IC50 of ∼ 3 µM, whereas BBI failed to suppress cell invasion, as measured by an in vitro invasion assay; 5)KTI suppresses HRA cell invasion by blocking uPA up-regulation which may be mediated by a binding protein(s) other thana bikunin binding protein and/or its receptor; and 6) that transforming growth factor-beta 1 (TGF-β1)-mediated activationof ERK1/2 was significantly reduced by preincubationof the cells with KTI. In conclusion, KTI, but not BBI, could inhibitcell invasiveness at least through suppression of uPA signaling cascade, although the mechanisms of KTI may be differentfrom those of bikunin.

Abbreviations: ELISA – enzyme-linked immunosorbent assay; ERK – extracellular signal-regulated kinase; JNK – c-Jun N-terminal kinase; MAP – mitogen-activated protein; MEK – mitogen-activated protein kinase/extracellular signal-regulatedkinase kinase; MMP – matrix metalloproteinase; PA(s) – plasminogen activator(s); TGF-β – transforming growth factor-β;uPA – urokinase-type plasminogen activator; uPAR – urokinase-type plasminogen activator receptor

Introduction

Invasive tumor cells not only express cell-associated pro-teases but also secrete anti-proteases, preventing the over-digestion of the extracellular matrix, which leads to a lossof cell attachment. The balance between proteolytic activ-ity and inhibition is crucial in the invasive and metastaticevent [1]. Therefore, drugs that manipulate or suppresssignal transduction on the plasminogen activator (PA) sys-tem in malignant cells offer a potentially new approach toanti-cancer therapy.

Bikunin, a Kunitz-type protease inhibitor, found in hu-man serum, amniotic fluid and urine is a serine proteaseinhibitor with antimetastatic activity [2–4]. It was sub-sequently demonstrated that bikunin inhibits tumor invasionand metastasis, at least in part, by a direct inhibition ofplasmin activity as well as by inhibiting uPA [2, 3] and

Correspondence to: Hiroshi Kobayashi, Department of Obstetrics andGynecology, Hamamatsu University School of Medicine, Handayama 1-20-1, Hamamatsu, Shizuoka, 431-3192, Japan. Tel: +81-53-4352309; Fax:+81-53-4352308; E-mail: hirokoba@hama-med.ac.jp

uPAR [4] expression at the gene and protein levels. Inter-estingly, in cell-free solutions, bikunin does not effectivelyinhibit uPA activity. Mechanistic studies in several cell typesdemonstrated that bikunin interferes with an upstream tar-get(s) of selected MAP kinase signaling processes suchas phosphorylation of MEK and ERK, leading to overex-pression of uPA [3]. We have reported that in preclinicalstudies with several types of rodent and human malignantcells, bikunin has proven effective as an anti-invasive andanti-metastatic agent [2–4]. In recently completed clinicalstudies, bikunin has shown promise at stabilizing diseaseprogression in patients with advanced ovarian cancer [5].

Protein protease inhibitors are widely distributed in plantseeds [6]. The trypsin–chymotrypsin inhibitors from soy-beans and from chickpeas inhibit insect midgut proteases,supporting the hypothesis that protease inhibitors comprise abuilt-in defense mechanism of the seed against insects. Soy-bean proteins are widely used in human foods in a variety offorms, including infant formulas, flour, protein concentrates,protein isolates, soy sauces, textured soy fibers, and tofu [6].The soybean trypsin inhibitor (SBTI) contains the soybean

160 H. Kobayashi et al.

Kunitz trypsin inhibitor (KTI) and the Bowman–Birk trypsinand chymotrypsin inhibitor (BBI). Infants on soy formulaconsume about 10 mg of KTI plus BBI per day.

The biochemical analysis demonstrates a functional dif-ference between KTI and BBI. BBI is also a potent anti-carcinogenic agent that inhibits chemical carcinogen- andradiation-induced malignant transformation in vitro and sup-presses carcinogenesis in several organ systems and animalspecies [7–9]. Findings on the involvement of protease in-hibitors, such as BBI, in prevention of tumorigenesis suggesta possible positive contribution of the inhibitors to the nu-tritional value of plant seeds. Bikunin and BBI are also aneffective inhibitor of nephrotoxicity induced by the antican-cer drug cisplatin [10] and the antibiotic gentamicin [11],respectively. Bikunin and BBI do not cause side effects anddo affect antimetastatic [5–7] or antimicrobial activity [11],respectively.

Furthermore, feeding experiments on diets containingisolated SBTI caused tumor growth depression in rats andchicks, but induced enlargement of the pancreas in rats,chicks and mice but not in pigs, dogs, calves, monkeys andpresumably humans [12]. It should be emphasized that all ofthese adverse effects are seen when protease inhibitors arepresent in relatively high concentrations in the diet and maybe completely unrelated to the anticarcinogenic effects seenat low concentrations of BBI [12].

In our experimental system, we chose the human ovariancancer cell line HRA because these cells mediate plas-minogen activation primarily by uPA but produce very lowamounts of matrix metalloproteases (MMPs), and the cellinvasion is uPA-dependent [13]. We examined whether KTIand BBI can reduce cell invasion by modulating cell sur-face expression of uPA. We also investigated whether, likehuman bikunin, KTI and BBI can inhibit the TGF-β1-stimulated uPA upregulation through suppression of theMAP kinase-dependent signaling.

Materials and methods

Cell culture

The human ovarian cancer cell line HRA was grown andcultured as described previously [3]. SKOV-3 ovarian cancercells were obtained from AmericanType Culture Collection.These cells were maintained in RPMI 1640 medium (Invit-rogen Corp., Carlsbad, California) supplemented with 10%fetal bovine serum in a cell culture incubator (constantly setat 37 ◦C with 5% CO2).

Chemicals and reagents

KTI and BBI were purified from soybeans as describedpreviously [14]. A specific ELISA for uPA was obtainedfrom American Diagnostica (Greenwich, Connecticut). Pro-teolytic enzymes, protease substrates, and specific proteaseinhibitors were also obtained from American Diagnosticaand include high molecular weight recombinant uPA, syn-thetic chromogenic substrate of uPA (Spectrozyme UK),

Glu-type plasminogen, purified plasmin, chromogenic plas-min substrate (Spectrozyme PL). Rabbit polyclonal anti-bodies against the total and phosphorylated ERK1/2 werefrom Zymed Laboratories Inc. (San Francisco, California)and Calbiochem (San Diego, California), respectively. Poly-clonal antibodies that recognize the total and phosphorylated(activated) forms of JNK and p38 were from Santa CruzBiotechnology (Santa Cruz, California). Ultrapure naturalhuman TGF-β1 was from Genzyme (Cambridge, Massachu-setts). MEK inhibitor PD98059 was from New EnglandBioLabs (Beverly, Massachusetts). Matrigel was purchasedfrom Collaborative Research (Bedford, Massachusetts). PP2(Calbiochem), and genistein were from Sigma-Aldrich Ja-pan (Tokyo, Japan). Unless otherwise specified, all otherchemicals and reagents were of the highest purity and ob-tained from Sigma. Carrier-free Na125I were purchased fromAmersham Pharmacia Biotech, Tokyo, Japan.

Preparation of conditioned medium

Cells were seeded in 35- or 100-mm dishes and basically24 h later washed twice with phosphate-buffered saline andthen treated with several reagents, including SBTI or TGF-β1 in serum-free medium. The conditioned medium wasproduced by incubating the cells (∼ 90% confluent) for24 h in a serum-free medium. All samples were stored at−20 ◦C until use. In parallel, cells treated in the same con-dition in different dishes were harvested and counted usinga hemocytometer.

Quantitation of uPA level

The amount of uPA was measured in the cell-conditionedmedium using an Imubind uPA ELISA kit from AmericanDiagnostica. Lower detection levels were set at 10 pg/ml forthe uPA kit.

Detection of cell surface-associated uPA by colorimetricassay

HRA cells were seeded at 10000 cells/well in 96-wellplates in the maintenance medium and allowed to reach∼ 90% confluence. The cells were washed once withphosphate-buffered saline and analyzed in a reaction buffer(100 µl/well) containing 0.8 mM MgCl2 and 0.2 µg/ml leu-peptin. The reactions were initiated by the addition of thechromogenic amidolytic substrate specific for uPA (Spec-trozyme UK) to a final concentration of 0.2 mM in theabsence or presence of 1 mM amiloride or neutralizing anti-bodies to 10 µg/ml uPA. No amidolytic substrate was addedin the blank reactions. The photometric absorbance of thereaction mixtures at 405 nm was monitored at 37 ◦C overthe next 30 min. Trypan blue cell staining procedure wasperformed.

Isolation of cytoplasmic RNA and Northern blot analysis

The HRA cells were plated at a density of 106 cells/100-mm dish. When cells grown in monolayer reached an early

Soybean trypsin inhibitor suppresses ovarian cancer cell invasion 161

phase of confluence, total cellular RNA samples were isol-ated from cells using a Qiagen RNeasy kit (Qiagen Ltd,Valencia, California) according to the manufacturer’s pro-tocol. For Northern blot analysis, 20 µg of cytoplasmic RNAwas electrophoresed onto a formaldehyde and 1.0% agarosegel and blotted onto a nylon filter. After transfer, RNA wascross-linked onto the membranes by UV irradiation. Thenylon filter was hybridized with a 32P-labeled cDNA probein 50% formamide, 5× saline-sodium phosphate, EDTA,0.1% SDS, 5× Denhardt’s solution, and 100 µg/ml salmonsperm DNA at 42 ◦C for 20 h. Quantitative analysis was per-formed with a Bio-Rad model 620 video densitometer witha one-dimensional Analyst software package for Macintosh.The uPA cDNA probe was prepared as described previously[3, 13]. The β-actin probe was an XhoI-XhoI fragment ofpHFβA-1 (ATCC, Rockville, Maryland).

Cell growth assay

HRA cells were grown on a flat bottomed 96-well plate at10000 cells/well containing the complete medium including10% fetal bovine serum. Twenty-four hours later, cells weregrown in RPMI 1640 medium with or without several re-agents. After 1, 3, and 5 days of culture, cell proliferationwas assayed by the addition of 20 µg of the vital dye MTT(1 mg/ml; Sigma). The blue dye taken up by cells during a4-h culture was dissolved in 100 µl/well dimethyl sulfoxide,and its A490 nm was read on an automated microplate reader.A preliminary study by MTT assay showed that absorbancewas directly proportional to the number of cells.

Effects of pharmacological inhibitors on uPA expression

HRA cells were cultured and serum starved. Cells were in-cubated for 24 h with one of the following agents: 50 µMMEK inhibitor PD98059, 25 µM general tyrosine kinaseinhibitor genistein, or 10 µM selective Src tyrosine kinaseinhibitor PP2. Control cells received just vehicle (0.1–0.5%dimethyl sulfoxide for PD98059, genistein, and PP2).

Western blotting

Confirmation of the presence of ERK1/2 phosphorylationwas obtained by Western blot using each monoclonal anti-body. Briefly, samples were electrophoresed under nonre-ducing conditions and transferred to polyvinylidene diflu-oride membrane using a semidry transfer system. Nonspe-cific binding sites were blocked for 1 h in Tris-bufferedsaline, pH 7.6, containing 2% bovine serum albumin.Blots were incubated with 0.5 µg/ml primary antibody for1 h and with secondary antibody (goat anti-mouse/rabbithorseradish peroxidase-conjugated IgG diluted 1:5000 inblocking solution) for 1 h. After several washes, reactivebands were visualized by a chemiluminescence detection kit(Amersham Biosciences, Piscataway, New Jersey).

Cell surface binding

Bikunin and KTI (150 µg/100 µl) were each radiolabeledwith 1 mCi of Na125I in the presence of 10 µg of chloramine

T and 0.15 M sodium phosphate, pH 7.5, and after adding83 µg of sodium metabisulfite purified on a PD-10 column(Amersham Biosciences) equilibrated and eluted with 0.2%BSA in PBS [15]. Specific activities were typically 1.4–3.2 µCi/µg. For binding studies, 200 µl aliquots of 5 × 105

cells in binding buffer (medium containing 0.1% BSA plus0.2% sodium azide) were incubated with increasing con-centrations of 125I-bikunin or 125I-KTI in the presence orabsence of an excess (up to 100-fold) of bikunin or KTI,respectively. After 2 h at 4 ◦C, the cells were washed bycentrifugation twice with cold binding buffer and counted.For displacement assays, 200 µl aliquots of 5 × 105 cellsin binding buffer were treated with 1.0 nM 125I-bikunin anddiffering concentrations of competitors (bikunin, KTI andBBI). In a parallel experiment, 200 µl aliquots of 5 × 105

cells in binding buffer were treated with 1.0 nM 125I-KTIand differing concentrations of competitors (bikunin, KTIand BBI). After 2 h, the cells were washed, and the boundradiolabel was quantitated by dissolving cells in NaOH.Specific binding was quantified using a multiwell γ -counter.

Invasion assay

The ability of cells to migrate across a Matrigel barrier (in-vasion) was determined by the modified Boyden chambermethod [16, 17]. Briefly, HRA cells (105/chamber) wereadded to polyvinylpyrrolidone-free, 8-µm polycarbonate fil-ters coated with 50 µl of 50 µg/ml Matrigel and incubatedwith complete medium containing 0.1% bovine serum al-bumin for 36 h at 37 ◦C. NIH3T3 fibroblast-conditionedmedium was used as a chemoattractant, and serum-free me-dium containing 0.1% bovine serum albumin was used asa negative control. Filters were removed from the chambersand stained with hematoxylin. Cells were counted at a 100 ×magnification, and the mean numbers of cells/field in fiverandom fields were recorded. Duplicate filters were used,and the experiments were repeated three times.

Statistics

Data are expressed as the mean ± S.D. of at least threeindependent triplicate experiments. Statistical analysis wasperformed by one-way analysis of variance followed byStudent’s t-test. P < 0.05 was considered statisticallysignificant.

Results

KTI, but not BBI, specifically suppresses unstimulated andTGF-β1-stimulated uPA expression

We have previously shown that HRA cells produce primaryuPA, but not MMPs and invade through an uPA-dependentmechanism [13]: HRA cell invasion is mediated by uPA up-regulation possibly through TGF-β1-dependent MAP kinaseactivation.

KTI and BBI were purified from soybeans as describedpreviously [14]. Under nonreducing conditions, KTI and

162 H. Kobayashi et al.

Figure 1. KTI and BBI purified from soybeans: Western blot analysis.SBTI contains KTI and BBI. An aliquot (10 µg/lane) was subjected toSDS-polyacrylamide gel electrophoresis under non-reducing conditions.Lane 1, KTI (20 kDa), and lane 2, BBI (8 kDa).

BBI have an apparent molecular mass of 20 and 8 kDa,respectively (Figure 1). To assess the specificity of the ef-fect of KTI and BBI, the conditioned medium of HRA cellswas tested by ELISA for uPA. ELISA data revealed thatBBI does not suppress uPA secretion into the conditionedmedium (Figure 2A). In contrast, treatment of HRA cellswith KTI induced a suppression of endogenously secreteduPA with doses as low as 1 µM, with an IC50 of ∼ 2 µM(Figure 2A). Confluent monolayers of HRA cells treatedwith KTI and BBI showed no evidence of cytotoxicity asmeasured by the trypan blue exclusion test or by the MTTassay (data not shown). Furthermore, addition of both KTIand BBI to the ELISA system does not directly affect theuPA level (not shown). We previously reported that TGF-β1 significantly stimulates uPA upregulation in HRA cells.Here we show that treatment of cells with KTI, but not BBI,specifically induced a suppression of the TGF-β1-stimulatedsecretion of uPA in a dose-dependent manner, with an IC50of ∼ 2 µM (Figure 2B).

Our previous data [13] demonstrated that the conditionedmedium of HRA cells was not very active in converting Glu-plasminogen to plasmin, which in turn did not lead to thecleavage of Spectrozyme PL. In contrast, when SpectrozymeUK was added to the monolayer of HRA cells, the Spec-trozyme UK was specifically cleaved. Amiloride (a catalyticinhibitor of uPA) 1 mM, and 10 µg/ml uPA-neutralizing an-tibody, but not 10 µg/ml tPA-neutralizing antibody, blockedthe cleavage of Spectrozyme UK on the plasma membraneof the HRA cells. It is worth noting that throughout thisbiochemical assay, cells remained intact, showing little orno plasma membrane permeability by trypan blue and a fullrecovery of proliferation in the subsequent cell growth assay(data not shown).

To determine whether the cell-associated uPA is affectedby exogenous SBTI, monolayer of HRA cells was pre-treated with KTI (Figure 3A) or BBI (Figure 3B) for 24 hand analyzed by colorimetric assay in the presence of theSpectrozyme UK. KTI, but not BBI, specifically suppresses

cell-associated uPA activity, irrespective of whether cellswere treated with TGF-β1. Figure 3 shows the results fromcells stimulated with TGF-β1. The alternative colorimetricassay confirmed that KTI and BBI at the concentrations usedin this study do not directly inhibit uPA catalytic activity in acell-free system (data not shown). These results demonstratethat KTI is able to inhibit uPA protein expression on theHRA cell surface.

KTI specifically inhibits expression of uPA mRNA in HRAcells treated with or without TGF-β1

TGF-β1 increased uPA mRNA expression starting 3 h afterstimulation, and the peak effect was observed at 6 h. No ef-fects were observed at the earlier 1-h time point. Further, thiseffect on uPA up-regulation was mediated through a TGF-βreceptor activation, since it was completely canceled by aneutralizing anti-TGF-β receptor antibody or TGF-β scav-enger, decorin (unpublished data [13]). Confluent cultures ofHRA cells were incubated in the presence of different con-centrations of KTI or BBI. As shown in Figure 4, Northernblot analysis of HRA cells incubated for 6 h in the presenceof 3 µM KTI revealed a significant decrease in uPA mRNAlevel (lanes 2 and 5). In contrast, BBI had no significanteffect on expression of uPA mRNA (lanes 3 and 6). No mor-phologically identifiable signs of cytotoxicity were observedin cells incubated with KTI or BBI.

Analysis of the signals regulating KTI-induced suppressionof TGF-β1-induced uPA expression

In the previous experiments, we showed that, when HRAcells were exposed to TGF-β1 for various periods, the phos-phorylation of ERK1/2 was enhanced from 5 min afteraddition of 10 ng/ml TGF-β1 and peaked at 15 min. In con-trast, JNK and p38 did not appear to be activated by TGF-β1.Thus, activation of ERK1/2 was most relevant among threesubsets of MAP kinase family members in response to TGF-β1. Furthermore, uPA upregulation is mediated by ERK1/2activation. Taken together TGF-β1 induces uPA upregula-tion through ERK1/2-dependent signaling cascade in HRAcells.

To analyze the signals regulating KTI-mediated suppres-sion of uPA, cells were treated with protein tyrosine kinaseinhibitor (genistein (25 µM), PP2 (10 µM)), MAP kinaseinhibitor (PD98059 (50 µM)), KTI (0.3 and 3 µM), or BBI(0.3 and 3 µM). Aspirate cell-conditioned medium was as-sayed for uPA levels by ELISA (Figure 5). The constitutiveand TGF-β1-stimulated expression of uPA observed in HRAcells was inhibited by preincubation of the cells with KTI,genistein, PP2, or PD98059, but not with BBI.

When HRA cells were incubated in the presence ofPD98059, TGF-β1-stimulated phosphorylation of ERK1/2was abrogated (Figure 6). In addition, the TGF-β1-stimulated ERK1/2 activation was also abrogated whenHRA cells were treated with KTI, but not BBI. Theseresults suggested that KTI specifically inhibits TGF-β1-dependent uPA expression possibly through suppression ofTGF-β1-mediated ERK1/2 activation.

Soybean trypsin inhibitor suppresses ovarian cancer cell invasion 163

Figure 2. Effects of KTI and BBI on uPA secretion HRA cells were seeded in 35-mm dishes and then treated with the indicated concentrations of KTI(white column) or BBI (black column) in the absence (A) or presence (B) of 10 ng/ml TGF-β1 for 24 h. The uPA contents of cell-conditioned mediumwere measured by ELISA. Results represent mean values ± S.D. of three experiments and triplicate determination. ∗P < 0.05 compared with the control.

Figure 3. Effects of KTI and BBI on cell-associated uPA activity. The monolayer of HRA cells seeded in 96-well plate was pretreated with KTI (0–10 µM,A) or BBI (0–10 µM, B) in the presence of 10 ng/ml TGF-β1 for 12 h. The monolayers were washed twice with PBS and analyzed by colorimetric assayin the presence of the Spectrozyme UK. Results represent mean values ± S.D. of three experiments and triplicate determination. ∗P < 0.05 comparedwith the control.

KTI, but not BBI, specifically suppressesTGF-β1-stimulated HRA cell invasion

Cell invasiveness was inhibited by treatment of the TGF-β1-stimulted cells with genistein (25 µM), PP2 (10 µM),PD98059 (50 µM), KTI (3 µM), whereas BBI (3 µM) failedto suppress cell invasion (Figure 7). We found that KTI couldspecifically reduce the invasion of the HRA cells treatedwith (Figure 7) or without TGF-β1 (data not shown). Thus,KTI specifically inhibited tumor cell invasion through sup-pression of endogenous and exogenous TGF-β1-mediatedexpression of uPA at the mRNA and protein level.

Assessment of generality of the results from HRA cells

In a parallel experiment, we have also used the TGF-β-sensitive SKOV-3 ovarian cancer cells on regulation of uPAexpression and invasion by SBTI. The key experiments wererepeated with HRA and SKOV-3 ovarian cancer cell lines.We confirmed that KTI, but not BBI, suppresses SKOV-3cell invasion by blocking uPA expression. These findingshave important implications for our understanding of therole of KTI in repressing TGF-β-mediated uPA expressionin ovarian cancer cells.

164 H. Kobayashi et al.

Figure 4. KTI, but not BBI, specifically inhibits expression of uPA mRNA.Upper panel: 106 HRA cells were plated in 100-mm dishes and culturedfor 48 h, then washed three times with phosphate-buffered saline. Confluentmonolayers of HRA cells were incubated for 6 h with the indicated concen-trations of KTI or BBI. Total RNAs were extracted and hybridized with uPAand β-actin cDNA. Each filter was loaded with 20 µg of total RNAs. Filterswere exposed for 24 h. Lower panel: data from three experiments werescanned and analyzed for quantification with the Macintosh software. Bandintensities for uPA were normalized to β-actin and presented as the mean± S.D. Results are from at least three separate experiments. ∗P < 0.05 vscontrol (lane 1); #P < 0.05 vs lane 4.

Direct binding and competitive inhibition of 125I-KTI or125I-bikunin to HRA cells

We tested direct binding of 125I-KTI to HRA cells. Fig-ure 8A shows that 125I-KTI and 125I-bikunin bound to thecells; half-maximal binding occurred at approximately 600and 400 nM, respectively. To characterize the binding of KTIto the cells, possibly via bikunin binding protein and/or itsreceptor, radiolabeled KTI was displaced by unlabeled com-petitors from each cell. As shown in Figure 8B, unlabeledKTI specifically displaced the binding of 1 nM 125I-KTIto the cells. The EC50, the concentration of unlabeled KTInecessary for 50% displacement, was ∼ 2 nM for HRAcells. Thus, binding of 125I-KTI was specific. In contrast,bikunin and BBI had no significant effect. Furthermore,we showed that the binding to HRA of 125I-bikunin wasblocked in the presence of unlabeled bikunin, but not KTI orBBI (Figure 8C). There may be a difference in the bindingmechanism between KTI and bikunin.

Discussion

Our previous data [2–4, 13] provide evidence that a Kunitz-type protease inhibitor, bikunin, not only has the capacityof functioning as a protease inhibitor to suppress directlycell surface-associated plasmin activity but it can also re-duce cell invasion by modulating uPA mRNA and protein

Figure 5. Effects of protein tyrosine kinase inhibitor, MAP kinase inhibitor,and SBTI (KTI and BBI) on the secretion of uPA. HRA cells were seeded in35-mm dishes and then treated with the indicated concentrations of drugs inserum-free medium for 24 h. The uPA contents of cell-conditioned mediumwere measured by ELISA. Results are the mean ± S.D. of three differentdeterminations. ∗P < 0.05 vs lane 2.

expression through inhibition of an upstream target(s) ofprotein kinase C- and/or MAP kinase-dependent signalingcascade [3]. More recent data [4] demonstrate that uPARexpression is also down-regulated by bikunin in humanchondrosarcoma HCS-2/8 cells and two human ovarian can-cer cell lines, HRA and HOC-I. Furthermore, we previouslyshowed that expression of the PA system in HRA cells isdependent on endogenously secreted TGF-β1, that exogen-ously added TGF-β1 can specifically induce uPA secretion,and that exogenously added TGF-β1 stimulated activationof MAP kinase (i.e. phosphorylation of ERK1/2), whichresulted in upregulation of uPA leading to enhancement ofHRA cell invasion [13]. In addition, expression of uPAwas inhibited by a protein tyrosine kinase inhibitor, gen-istein (25 µM), and a selective inhibitor of Src tyrosinekinase, PP2 (10 µM). PP2 also prevented the activation ofERK1/2 [13]. Taken together, TGF-β1 can stimulate uPAexpression through the Src-dependent, ERK-specific sig-naling cascade (J. Biol. Chem. 2003, submitted). Further,TGF-β1-induced overexpression of uPAR [18] may leadto activation of MAP kinase components, demonstrating across-talk between TGF-β1-mediated MAP kinase cascadeand the uPA/uPAR system.

Here we show that, like human bikunin, KTI purifiedfrom soybeans may also function at the upstream target ofMAP kinase cascade. This is in agreement with previousdata that the ERK cascade, often an event downstream of Srctyrosine kinase activation, has been shown to mediate uPAexpression to a variety of agents, including phorbol ester [19,20] and that bikunin could inhibit an upstream componentof the ERK cascade involved in phorbol ester-mediated uPAexpression in human HRA cells [3, 4, 13] and human chon-

Soybean trypsin inhibitor suppresses ovarian cancer cell invasion 165

Figure 6. Effects of SBTI (KTI and BBI) on the TGF-β1-stimulated ac-tivation of the ERK1/2 phosphorylation. Upper panel: HRA cells wereexposed to 10 ng/ml TGF-β1 in the absence or presence of KTI or BBIfor 15 min. Total and phosphorylated ERK1/2 were measured by West-ern blotting. Lower panel: data from three experiments were scanned andanalyzed for quantification with the Macintosh software. Band intensitiesfor phospho-Akt were normalized to total Akt and presented as the mean± S.D. Results are from at least two separate experiments with similarresults. ∗P < 0.05 vs lane 2.

drosarcoma cell line HCS-2/8 [3]. The functional differencesbetween KTI and BBI on suppression of uPA upregulationmay be the result of variability in drug metabolism or avariability of the amounts of specific receptors for SBTI ontumor cell membrane.

It has been reported that there appeared to be a specificinteraction between bikunin and the tumor cell surface. Tu-mor cells expresses two types of bikunin-binding proteins(BPs); a 40-kDa bikunin-BP, which is identical to cartilagelink protein, and a 45-kDa bikunin-BP, a putative bikunin re-ceptor [21]. HRA cells do not produce endogenous bikunin,but they do exhibit two types of bikunin receptors [21].Bikunin at low concentrations (< 200 nM) bound to HRAcells in a dose-dependent manner. As the concentration ofbikunin exceeded 200 nM, the amount of cell surface-boundbikunin was not further increased, suggesting that the bind-ing of bikunin had reached a level of saturation (data notshown). Bikunin binds link protein and bikunin receptor ontumor cell surface possibly via the Kunitz domain I and thechondroitin sulfate side chain, respectively. Bikunin mustbind directly to both of the cell-associated bikunin-BPs tosuppress expression of uPA and uPAR genes.

Our preliminary data showed that neither KTI not BBIinhibits binding of radiolabeled-bikunin to HRA cells orbikunin does not inhibit KTI binding to the cells (Fig-ure 8). KTI suppresses ovarian cancer cell invasion byblocking uPA up-regulation which is mediated by a bind-ing protein(s) other than a bikunin binding protein and/orits receptor, suggesting that their binding molecule(s) inthe plasma membranes are different from those of bikunin.These findings lead up many questions about the differencein the mechanisms between KTI and bikunin, because KTI

Figure 7. Effects of pharmacological inhibitors and SBTI (KTI and BBI)on cell invasion. A cell suspension, 200 µl (500,000 cells/ml) in RPMI1640 containing 1% fetal bovine serum, was placed on the Matrigel-coatedsurface of the upper chamber. The indicated concentrations of genistein,PP2, PD98059, KTI or BBI in the presence of 10 ng/ml TGF-β1 wereadded to the upper compartment. After a 36-h incubation, cells that invadedthrough the Matrigel-coated membrane were stained and counted under themicroscope. All experiments were performed three times, and typical dataare shown. Results are the mean ± S.D. of three different determinations.∗P < 0.05 vs lane 2.

and bikunin are members of the Kunitz family and both KTIand bikunin show suppression of uPA expression in a sim-ilar fashion. In addition, the part of the bikunin responsiblefor cell binding is both the N-terminal Kunitz-domain andchondroitin-4-sulfate side chain of bikunin. However, KTIhas no chondroitin-4-sulfate glycosaminoglycan side chain.These data allow us to hypothesize that the mechanisms ofKTI will be different from those of bikunin. Further exper-iments are needed in order to clarify whether HRA cellsexpress specific receptors or binding sites for KTI or BBI.

SBTI was shown previously to have anti-carcinogenicproperties at low micromolar concentrations [22]. The BBIfound in soybeans is a serine protease inhibitor with anti-carcinogenic activity [22]: BBI can also effectively inhibitN-nitrosomethylbenylamine (NMBzA)-induced esophagealtumors [22]. SBTI showed significant reductions in colla-genolytic activity [23]. These data allow us to speculatethat KTI and BBI purified from soybeans may affect theantimetastatic and anticarcinogenic activities, respectively.

We report here, for the first time, the potent ability ofKTI from soybeans to downregulate uPA expression andsubsequently inhibit invasion in ovarian cancer cells pos-sibly through suppression of TGF-β1-dependent ERK1/2activation. At this time, however, we did not show evidenceestablishing that the functional difference of these inhibitorsmay be a result of differences in either drug metabolismor in surface receptor expression, or that these differencesmay result in differing anti-metastatic or anti-carcinogenicactivities for these inhibitors.

166 H. Kobayashi et al.

Figure 8. (A) Direct binding and competitive inhibition of 125I-KTI or 125I-bikunin to HRA cells A, direct binding of 125I-KTI or 125I-bikunin to HRAcells. Maximum specific binding of 125I-KTI (n = 3) or 125I-bikunin (n = 3) was considered to be 100%. (B) Displacement of the binding of 125I-KTIto the cells. (C) Displacement of the binding of 125I-bikunin to the cells. Open circles, bikunin; filled circles, KTI; open squares, BBI. Results arerepresentative of three individual experiments. A typical binding curves (A) and displacement curves (B, C) were shown.

Acknowledgements

The authors thank Drs Ryuji Yoshida, Yasufumi Kanada,Setsuko Nishiyama, and Youichi Fukuda (Fuji Oil Co., Ltd,Tsukuba R&D Center, 4–3 Kinunodai, Yawara, Tsukuba-gun, Ibaraki, 300-2497, Japan) for their continuous andgenerous support of our work. This work was supported bya grant-in-aid for Scientific Research from the Ministry ofEducation, Science, and Culture of Japan (to H. K.) and bya grant from the Fuji Foundation for Protein Research (toH.K.).

References

1. Reuning U, Magdolen V, Wilhelm O et al. Multifunctional potentialof the plasminogen activation system in tumor invasion and metastasis(Review). Int J Oncol 1998; 13 (5): 893–906.

2. Kobayashi H, Shinohara H, Gotoh J et al. Anti-metastatic therapy byurinary trypsin inhibitor in combination with an anti-cancer agent. BrJ Cancer. 1995; 72 (5): 1131–7.

3. Kobayashi H, Suzuki M, Tanaka Y et al. Suppression of urok-inase expression and invasiveness by urinary trypsin inhibitor ismediated through inhibition of protein kinase C- and MEK/ERK/c-Jun-dependent signaling pathways. J Biol Chem 2001; 276 (3):2015–22.

4. Kobayashi H, Suzuki M, Kanayama N et al. Suppression of urok-inase receptor expression by bikunin is associated with inhibition ofupstream targets of extracellular signal-regulated kinase-dependentcascade. Eur J Biochem 2002; 269 (16): 3945–57.

5. Kobayashi H, Suzuki M, Hirashima Y et al. The protease inhibitorbikunin, a novel anti-metastatic agent. Biol Chem. 2003; 384 (5):749–54.

6. Brandon DL, Bates AH, Friedman M. ELISA analysis of soybeantrypsin inhibitors in processed foods. Adv Exp Med Biol 1991; 289:321–37.

7. Kennedy AR, Troll W, Kennedy AR (eds). Protease Inhibitors as Can-cer Chemopreventive Agents. New York: Plenum Publishing 1993;65–91.

8. Kennedy AR, Troll W, Kennedy AR (eds). Protease Inhibitors as Can-cer Chemopreventive Agents. New York: Plenum Publishing 1993;9–64.

9. Kennedy AR. Chemopreventive agents: protease inhibitors. Phar-macol Ther 1998; 78 (3): 167–209.

10. Kobayashi H, Ohi H, Terao T. Prevention by urinastatin of cis-diamminedichloroplatinum-induced nephrotoxicity in rabbits: Com-parison of urinary enzyme excretions and morphological alterationsby electron microscopy. Asia Oceania J Obstet Gynaecol 1991; 17(3): 277–88.

11. Birk Y. Protein proteinase inhibitors in legume seeds – overview. ArchLatinoam Nutr 1996; 44 (4 Suppl 1): 26S–30S.

12. Liener IE. Possible adverse effects of soybean anticarcinogens. J Nutr1995; 125 (3 Suppl): 744S–750S.

13. Kobayashi H, Suzuki M, Tanaka Y et al. A Kunitz-type proteaseinhibitor, bikunin, inhibits ovarian cancer cell invasion by blockingthe calcium-dependent transforming growth factor-beta 1 signalingcascade. J Biol Chem 2003; 278 (10): 7790–9.

14. Pusztai A, Watt WB, Steward JC. A comprehensive scheme for theisolation of trypsin inhibitors and the agglutinin from soybean seeds.J Agric Food Chem 1991; 39: 882–6.

15. Morrissey D, O’Connell J, Lynch D et al. Invasion by esophagealcancer cells: functional contribution of the urokinase plasminogenactivation system, and inhibition by antisense oligonucleotides tourokinase or urokinase receptor. Clin Exp Metast 1999; 17: 77–85.

16. Hirashima Y, Kobayashi H, Suzuki M et al. Transforming growthfactor-beta1 produced by ovarian cancer cell line HRA stimulatesattachment and invasion through an up-regulation of plasminogen ac-tivator inhibitor type-1 in human peritoneal mesothelial cells. J BiolChem 2003; 278 (29): 26793–802.

17. Albini A, Iwamoto Y, Kleinman HK et al. A rapid in vitro assay forquantitating the invasive potential of tumor cells. Cancer Res 1987;47 (12): 3239–45.

18. Shetty S, Idell S. Posttranscriptional regulation of urokinase receptorgene expression in human lung carcinoma and mesothelioma cells invitro. Mol Cell Biochem 1999; 199 (1–2): 189–200.

19. Bell SM, Connolly DC, Maihle NJ, Degen JL. Differential modula-tion of plasminogen activator gene expression by oncogene-encodedprotein tyrosine kinases. Mol Cell Biol 1993; 13 (9): 5888–97.

20. Bell SM, Brackenbury RW, Leslie ND, Degen JL. Plasminogen ac-tivator gene expression is induced by the src oncogene product andtumor promoters. J Biol Chem 1990; 265 (3): 1333–8.

21. Suzuki M, Kobayashi H, Tanaka Y et al. Structure and functionanalysis of urinary trypsin inhibitor (UTI): Identification of bind-ing domains and signaling property of UTI by analysis of truncatedproteins. Biochim Biophys Acta. 2001; 1547 (1): 26–36.

22. von Hofe E, Newberne PM, Kennedy AR. Inhibition of N-nitrosomethylbenzylamine-induced esophageal neoplasms by theBowman-Birk protease inhibitor. Carcinogenesis 1991; 12 (11):2147–50.

23. Lewin MR, Chowcat NL, Jayaraj AP, Boulos PB. Collagenase inhibi-tion in colonic mucosa by proteinase inhibitors. Br J Exp Pathol 1986;67 (4): 523–6.